2D students to top-level setup_53_2d/ setup_54_2d/ (3D manifest schema, per-axis sizes, step 20000); source_repo on 3D manifests; remove 2d/

2d/README.md

@@ -1,14 +0,0 @@
-# 2D-slice mito models (experimental)
-
-One z-slice in, one slice out: `input` float32 `[1,1,1,N,N]` → `output` float32 raw logits `[1,3,1,N-88,N-88]`
-(mito affinities z, y, x; threshold at 0). 16 nm, input norm `uint8/255*2-1`, same as the 3D models.
-
-These are **distilled students** of the 3D models `setup_53` / `setup_54`: every 3D kernel collapsed to 2D (sum of depth
-taps) and fine-tuned to reproduce the 3D model's predictions slice by slice (no labels; random axis permutations so
-xz/yz views work too). Collapsing alone gives F1 ≈ 0.08 (53) / 0.38 (54); distillation recovers most of the 3D accuracy.
-Checkpoint step and per-file fidelity (vs the PyTorch student) are in each folder's `manifest.json`.
-
-WebGPU (Chrome, Apple M-series, fp16): setup_53 1068² → 980² in 264 ms (3.6 Mpix/s), 556² → 468² in 64 ms,
-vs 752 ms per 172³ → 84³ block for the 3D model (0.79 Mvox/s).
-
-cellmap-flow serverless: Input size `1,N,N`, Output size `1,N-88,N-88`, voxel 16,16,16, Channels 3, Has channel axis true.

setup_53/manifest.json

@@ -177,5 +177,6 @@
       "flip_frac": 0.0001571787837219936,
       "ort_cpu_s": 0.76
     }
-  ]
+  ],
+  "source_repo": "cellmap/mito-aff-unetiso-setup-53"
 }

setup_53_2d/README.md

@@ -0,0 +1,10 @@
+# setup_53_2d — 2D-slice mito student (experimental)
+
+One z-slice in, one slice out: `input` float32 `[1,1,1,N,N]` → `output` float32 raw logits `[1,3,1,N-88,N-88]`
+(mito affinities z, y, x; threshold at 0). 16 nm, input norm `uint8/255*2-1`.
+
+Distilled from the 3D model `setup_53` (PyTorch: `cellmap/mito-aff-unetiso-setup-53`): every 3D kernel collapsed to 2D (sum of depth taps) and
+fine-tuned to reproduce the 3D predictions slice by slice. Training step, per-file fidelity and sizes are in
+`manifest.json` (same schema as the 3D folders, per-axis `input_size` / `output_size`).
+
+cellmap-flow serverless: Input size `1,N,N`, Output size `1,N-88,N-88`, voxel 16,16,16, Channels 3, Has channel axis true.

{2d/setup_53 → setup_53_2d}/manifest.json

@@ -1,15 +1,25 @@
 {
-  "model": "setup_53",
-  "kind": "2D slice student distilled from the 3D model (train_2d_distill.py)",
-  "checkpoint": "runs/setup_53_2d/student_20000.pt",
-  "step": 20000,
+  "model": "setup_53_2d",
+  "task": "mito_affinities",
+  "architecture": "Unet3dIso (2D-collapsed)",
+  "checkpoint": "student_20000",
+  "params": 1927907,
   "channels": [
     "mito_aff_z",
     "mito_aff_y",
     "mito_aff_x"
   ],
-  "voxel_size_nm": 16,
-  "io": "input float32 [1,1,1,N,N] (one z-slice); output float32 raw logits [1,3,1,N-88,N-88]",
+  "input_voxel_size_nm": [
+    16,
+    16,
+    16
+  ],
+  "output_voxel_size_nm": [
+    16,
+    16,
+    16
+  ],
+  "output": "raw logits (apply sigmoid / threshold at 0)",
   "input_norm": [
     {
       "name": "MinMaxNormalizer",
@@ -21,56 +31,115 @@
       "expression": "2*x-1"
     }
   ],
+  "io": "input 'input' float32 [1,1,1,y,x] (one z-slice); output 'output' float32 [1,C,1,y',x']",
+  "kind": "2D slice student distilled from the 3D model (train_2d_distill.py)",
+  "step": 20000,
+  "distilled_from": "setup_53",
+  "distilled_from_repo": "cellmap/mito-aff-unetiso-setup-53",
   "variants": [
     {
+      "model": "setup_53_2d",
       "variant": "fp32",
-      "input_size": 300,
-      "output_size": 212,
+      "input_size": [
+        1,
+        300,
+        300
+      ],
+      "output_size": [
+        1,
+        212,
+        212
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in300_fp32.onnx",
       "file_mb": 7.75,
       "max_abs": 6.103515625e-05,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_53_2d",
       "variant": "fp16",
-      "input_size": 300,
-      "output_size": 212,
+      "input_size": [
+        1,
+        300,
+        300
+      ],
+      "output_size": [
+        1,
+        212,
+        212
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in300_fp16.onnx",
       "file_mb": 3.9,
       "max_abs": 0.13983917236328125,
       "flip_frac": 5.1916459000830666e-05
     },
     {
+      "model": "setup_53_2d",
       "variant": "fp32",
-      "input_size": 556,
-      "output_size": 468,
+      "input_size": [
+        1,
+        556,
+        556
+      ],
+      "output_size": [
+        1,
+        468,
+        468
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in556_fp32.onnx",
       "file_mb": 7.75,
       "max_abs": 6.103515625e-05,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_53_2d",
       "variant": "fp16",
-      "input_size": 556,
-      "output_size": 468,
+      "input_size": [
+        1,
+        556,
+        556
+      ],
+      "output_size": [
+        1,
+        468,
+        468
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in556_fp16.onnx",
       "file_mb": 3.9,
       "max_abs": 0.15328407287597656,
       "flip_frac": 4.717900017045316e-05
     },
     {
+      "model": "setup_53_2d",
       "variant": "fp32",
-      "input_size": 1068,
-      "output_size": 980,
+      "input_size": [
+        1,
+        1068,
+        1068
+      ],
+      "output_size": [
+        1,
+        980,
+        980
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in1068_fp32.onnx",
       "file_mb": 7.75,
       "max_abs": 6.866455078125e-05,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_53_2d",
       "variant": "fp16",
-      "input_size": 1068,
-      "output_size": 980,
+      "input_size": [
+        1,
+        1068,
+        1068
+      ],
+      "output_size": [
+        1,
+        980,
+        980
+      ],
       "file": "setup_53_mito_affinities_16nm_2d_in1068_fp16.onnx",
       "file_mb": 3.9,
       "max_abs": 0.16084671020507812,

setup_54/manifest.json

@@ -177,5 +177,6 @@
       "flip_frac": 7.592322643343052e-05,
       "ort_cpu_s": 0.96
     }
-  ]
+  ],
+  "source_repo": "cellmap/mito-aff-resunet-setup-54"
 }

setup_54_2d/README.md

@@ -0,0 +1,10 @@
+# setup_54_2d — 2D-slice mito student (experimental)
+
+One z-slice in, one slice out: `input` float32 `[1,1,1,N,N]` → `output` float32 raw logits `[1,3,1,N-88,N-88]`
+(mito affinities z, y, x; threshold at 0). 16 nm, input norm `uint8/255*2-1`.
+
+Distilled from the 3D model `setup_54` (PyTorch: `cellmap/mito-aff-resunet-setup-54`): every 3D kernel collapsed to 2D (sum of depth taps) and
+fine-tuned to reproduce the 3D predictions slice by slice. Training step, per-file fidelity and sizes are in
+`manifest.json` (same schema as the 3D folders, per-axis `input_size` / `output_size`).
+
+cellmap-flow serverless: Input size `1,N,N`, Output size `1,N-88,N-88`, voxel 16,16,16, Channels 3, Has channel axis true.

{2d/setup_54 → setup_54_2d}/manifest.json

@@ -1,15 +1,25 @@
 {
-  "model": "setup_54",
-  "kind": "2D slice student distilled from the 3D model (train_2d_distill.py)",
-  "checkpoint": "runs/setup_54_2d/student_20000.pt",
-  "step": 20000,
+  "model": "setup_54_2d",
+  "task": "mito_affinities",
+  "architecture": "ResUNet3DValid (2D-collapsed)",
+  "checkpoint": "student_20000",
+  "params": 3750963,
   "channels": [
     "mito_aff_z",
     "mito_aff_y",
     "mito_aff_x"
   ],
-  "voxel_size_nm": 16,
-  "io": "input float32 [1,1,1,N,N] (one z-slice); output float32 raw logits [1,3,1,N-88,N-88]",
+  "input_voxel_size_nm": [
+    16,
+    16,
+    16
+  ],
+  "output_voxel_size_nm": [
+    16,
+    16,
+    16
+  ],
+  "output": "raw logits (apply sigmoid / threshold at 0)",
   "input_norm": [
     {
       "name": "MinMaxNormalizer",
@@ -21,56 +31,115 @@
       "expression": "2*x-1"
     }
   ],
+  "io": "input 'input' float32 [1,1,1,y,x] (one z-slice); output 'output' float32 [1,C,1,y',x']",
+  "kind": "2D slice student distilled from the 3D model (train_2d_distill.py)",
+  "step": 20000,
+  "distilled_from": "setup_54",
+  "distilled_from_repo": "cellmap/mito-aff-resunet-setup-54",
   "variants": [
     {
+      "model": "setup_54_2d",
       "variant": "fp32",
-      "input_size": 300,
-      "output_size": 212,
+      "input_size": [
+        1,
+        300,
+        300
+      ],
+      "output_size": [
+        1,
+        212,
+        212
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in300_fp32.onnx",
       "file_mb": 15.05,
       "max_abs": 1.9550323486328125e-05,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_54_2d",
       "variant": "fp16",
-      "input_size": 300,
-      "output_size": 212,
+      "input_size": [
+        1,
+        300,
+        300
+      ],
+      "output_size": [
+        1,
+        212,
+        212
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in300_fp16.onnx",
       "file_mb": 7.56,
       "max_abs": 0.012979507446289062,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_54_2d",
       "variant": "fp32",
-      "input_size": 556,
-      "output_size": 468,
+      "input_size": [
+        1,
+        556,
+        556
+      ],
+      "output_size": [
+        1,
+        468,
+        468
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in556_fp32.onnx",
       "file_mb": 15.05,
       "max_abs": 4.76837158203125e-05,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_54_2d",
       "variant": "fp16",
-      "input_size": 556,
-      "output_size": 468,
+      "input_size": [
+        1,
+        556,
+        556
+      ],
+      "output_size": [
+        1,
+        468,
+        468
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in556_fp16.onnx",
       "file_mb": 7.56,
       "max_abs": 0.024660348892211914,
       "flip_frac": 1.2175225850439526e-05
     },
     {
+      "model": "setup_54_2d",
       "variant": "fp32",
-      "input_size": 1060,
-      "output_size": 972,
+      "input_size": [
+        1,
+        1060,
+        1060
+      ],
+      "output_size": [
+        1,
+        972,
+        972
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in1060_fp32.onnx",
       "file_mb": 15.05,
       "max_abs": 0.0002117156982421875,
       "flip_frac": 0.0
     },
     {
+      "model": "setup_54_2d",
       "variant": "fp16",
-      "input_size": 1060,
-      "output_size": 972,
+      "input_size": [
+        1,
+        1060,
+        1060
+      ],
+      "output_size": [
+        1,
+        972,
+        972
+      ],
       "file": "setup_54_mito_affinities_16nm_2d_in1060_fp16.onnx",
       "file_mb": 7.56,
       "max_abs": 0.02630472183227539,

setup_55/manifest.json

@@ -175,5 +175,6 @@
       "flip_frac": 0.0,
       "ort_cpu_s": 0.72
     }
-  ]
+  ],
+  "source_repo": "cellmap/nuc-unetiso-setup-55"
 }

setup_56/manifest.json

@@ -175,5 +175,6 @@
       "flip_frac": 0.0006377551020408163,
       "ort_cpu_s": 0.88
     }
-  ]
+  ],
+  "source_repo": "cellmap/nuc-resunet-setup-56"
 }

setup_61/manifest.json

@@ -175,5 +175,6 @@
       "flip_frac": 0.0,
       "ort_cpu_s": 0.98
     }
-  ]
+  ],
+  "source_repo": "cellmap/nuc-resunet-setup-61"
 }

setup_67/manifest.json

@@ -175,5 +175,6 @@
       "flip_frac": 0.002824344023323615,
       "ort_cpu_s": 0.88
     }
-  ]
+  ],
+  "source_repo": "cellmap/er-resunet-setup-67"
 }